基于视觉惯性里程计与语义信息的无人机SLAM方法研究

Research on UAV SLAM method based on visual inertial odometer and semantic information

室内环境中存在丰富的语义信息,可以使机器人更好地理解环境,提高机器人位姿估计的准确性。虽然语义信息在机器人同时定位与地图构建(SLAM)领域得到了深入研究和广泛应用,但是在环境准确感知、语义特征提取和语义信息利用等方面还存在着很多困难。针对上述难点,提出了一种基于视觉惯性里程计算法与语义信息相结合的新方法,该方法通过视觉惯性里程计来估计机器人的状态,通过校正估计,构建从语义检测中提取的几何表面的稀疏语义地图;通过将检测到的语义对象的几何信息与先前映射的语义信息相关联来解决视觉惯性里程计和惯性测量单元的累积误差问题。在室内环境中对装备RGB-D深度视觉和激光雷达的无人机进行验证实验,结果表明,该方法比视觉惯性里程计算法取得了更好的结果。应用结合语义信息和视觉惯性里程计的SLAM算法表现出很好的鲁棒性和准确性,该方法能提高无人机导航精度,实现无人机智能自主导航。

There are abundant semantic information in the indoor environment, which can make the robot better understand the environment and improve the accuracy of robot pose estimation. Although semantic information has been deeply studied and widely used in the field of robot simultaneous localization and map construction(SLAM), there are still many difficulties in accurate environment perception, semantic feature extraction and semantic information utilization. In view of the above difficulties, this paper proposed a new method based on the combination of VIO algorithm and semantic information. This method estimated the state of the robot through VIO, and constructed the sparse semantic map of the geometric surface extracted from semantic detection through correction estimation. It solved the cumulative error of VIO and inertial measurement unit by associating the geometric information of the detected semantic object with the previously mapped semantic information. The experimental results show that this method obtains better results than the VIO algorithm. The SLAM algorithm combining semantic information and VIO shows the advantages of good robustness and accuracy. This method improves the navigation accuracy of UAV, realizes UAV intelligent autonomous navigation.